Ring-Amplification Technique for Bio-Signal LNA Designs
Zhiyang Song
David Allstot, Ed.
Electrical Engineering and Computer Sciences
University of California at Berkeley
Technical Report No. UCB/EECS-2015-82
http://www.eecs.berkeley.edu/Pubs/TechRpts/2015/EECS-2015-82.html
May 13, 2015
Copyright © 2015, by the author(s).
All rights reserved.
Permission to make digital or hard copies of all or part of this work for
personal or classroom use is granted without fee provided that copies are
not made or distributed for profit or commercial advantage and that
copies bear this notice and the full citation on the first page. To copy
otherwise, to republish, to post on servers or to redistribute to lists,
requires prior specific permission.
Ring-Amplification Technique for Bio-Signal LNA Designs
by
Zhiyang Song
A capstone Report Submitted in Partial Fulfillment of
the Requirements for the Degree of
Master of Engineering
in
Electrical Engineering and Computer Sciences
in the
Graduate Division
of the
University of California, Berkeley
Review Committee in Charge:
Professor David J. Allstot, Capstone Advisor
Professor Jan M. Rabaey
Spring 2015
University of California, Berkeley College of Engineering
MASTER OF ENGINEERING - SPRING 2015
Electrical Engineering and Computer Sciences
Integrated Circuits
Ring-Amplification Technique for Bio-Signal LNA Designs
Zhiyang Song
This Masters Project Paper fulfills the Master of Engineering degree
requirement.
Approved by:
1. Capstone Project Advisor:
Signature: __________________________ Date ____________
Print Name/Department:
David J. Allstot/Electrical Engineering and Computer Sciences
2. Faculty Committee Member #2:
Signature: __________________________ Date ____________
Print Name/Department:
Jan M. Rabaey/Electrical Engineering and Computer Sciences
1
Abstract
Low power wireless medical electronic design is a heated research topic due to the
development of remote medical diagnosis systems and long-term treatment assistances. In
general, such medical devices contain bio-signal sensor that monitor the health status of a patient
or health-conscious person. Our main goal is to design an LNA that is suitable for amplifying
bio-signals in low-power remote sensing devices. In this work, the business strategy and the IP
strategy is being studied and presented.
Ring-amplification technique is a novel amplifier design approach, which has the
potential to achieve high gain with minimal power dissipation. The motif behind applying ringamplification technique to low-power LNA design is that ring-amplifier has the potential to
achieve high gain with low power dissipation as well as a relatively simple structure. This work
reviews the current status of ring-amplifier design, as well as explores the ring-amplifier design
process. The original ring-amplifier approach is not suitable for low-noise amplifier design due
to its single-ended structure, therefore effort is being made to modify the ring-amplifier structure
to make the amplifier suitable for employing flicker noise cancellation techniques. The eventual
result of this work, the differential input ring-amplifier employing chopper stabilization, which
achieved closed-loop gain of 63.93dB, THD -78dB, UGB of 12MHz and a positive phase margin
which indicates the stability of the ring-amplifier. At 10Hz the output noise voltage density is
120uV/(Hz)1/2, which is approximately an order of magnitude noise reduction at low frequencies,
compared to the simple ring-amplifier.
Keywords: Bio-signal; Low Noise Amplifier; Ring Amplifier; Low Power; Low Frequency.
2
Table of Contents
1. Problem Statement ......................................................................................................................... 4 2. Business Strategy for Low-­‐Power LNA (Low-­‐Noise Amplifier) ........................................ 5 2.1. Introduction .......................................................................................................................... 5
2.2. Identification of Industry ..................................................................................................... 6
2.3. Potential Impact to the Industry ........................................................................................... 7
2.4. Threat of Alternative Technology ........................................................................................ 8
2.5. Competition in the Industry ................................................................................................. 9
2.6. Potential Buyers, Suppliers, and End Users....................................................................... 10
2.7. Power of Buyers and Suppliers .......................................................................................... 11
2.8. Basic Marketing Strategy ................................................................................................... 12
2.9. Social Trends ..................................................................................................................... 13
2.10. Technological Trend ........................................................................................................ 13
2.11. Economic Trends ............................................................................................................. 14
2.12. Regulatory Trends............................................................................................................ 15
2.13. Business Strategy to Cope with Contemporary Trends ................................................... 15
3. Intellectual Property Strategy for Ultra-­‐low Power LNA ................................................. 17 3.1. Introduction ........................................................................................................................ 17
3.2. Patentability of Our Design ............................................................................................... 17
3.3. Existing Similar Patent ...................................................................................................... 18
3.4. Possible Similarity and Distinction.................................................................................... 19
3.5. Alternative Options and Potential Risks ............................................................................ 20
4. Technical Contributions .............................................................................................................. 22 4.1. Overview ............................................................................................................................ 22
4.2. Literature Review............................................................................................................... 23
4.2.1. Common-Source Fully-Differential Amplifier ........................................................... 24
4.2.2. Ring-Amplifier............................................................................................................ 25
4.3. Methods and Materials ....................................................................................................... 29
4.3.1. Ring Amplifier Design Process .................................................................................. 29
4.3.2. Adjustments and Improvement ................................................................................... 33
4.4. Results and Discussion ...................................................................................................... 37
5. Concluding Reflections ................................................................................................................ 47 3
1. Problem Statement
Low power wireless medical devices are currently in the spotlight due to the development
of remote medical diagnosis systems and long-term treatment assistances. In general, such
medical devices contain bio-signal sensor that monitors the health status of a patient or healthconscious person. The output of the sensor is extremely weak for further processing and
susceptible to noise. Thus, a low-noise amplifier is connected to the sensor to amplify the
original input signal while minimizing the effect of noise. The specification of a low-noise
amplifier (LNA) determines the quality of the processed signal.
Our main goal is to design an LNA that is suitable for implantable bio-signal acquisition
systems. We primarily focus on the implantable medical system monitoring the electrical signal
of a person’s heart. The LNA provides enough voltage gain to amplify the signals while reducing
the noise contamination significantly.
In addition, its power consumption is minimized to enhance the lifespan of the whole
system. Currently, a bio-signal acquisition system’s overall power consumption is dominated by
the low-noise amplifier. The other module in the bio-signal acquisition system only consumes
less than 1uW each, while a typical LNA consumes more than 10uW. By reducing the power
consumption of the LNA, the device can maintain its performance or longer period of time.
4
2. Business Strategy for Low-Power LNA (Low-Noise Amplifier)
2.1. Introduction
According to Deaths: Final Data for 2013, a recent statistical report issued by Centers
for Disease Control and Prevention, life expectancy of people at birth in the United States in
1950 is only 68.2 years (Center for Disease Controls and Prevention). In contrast, life expectancy
at birth in 2013 is 78.8 years. Nowadays, people are expected to live almost 10 years longer than
people living the middle of 1990s. Although life expectancy at birth has been steadily increasing
for the past years, people are still exposed to danger of chronic diseases, such as diabetes,
hypertension, and arrhythmia.
Deaths: Final Data for 2013 shows that the leading cause of death in 2013 is diseases in
heart. Moreover, diabetes is the seventh leading cause of death (Center for Disease Controls and
Prevention). These chronic diseases have relevant bio-signals that indicate health status of
patients. For instance, blood sugar level of a person who suffers in diabetes indicates when the
person needs to inject insulin. If real-time monitoring of bio-signals is possible, doctors are able
to anticipate urgent situations so that more lives can be saved. Recently, wearable or implantable
wireless medical devices are developed to detect bio-signals from a patient and send data
remotely to other devices.
The goal of our Capstone Project is to design a low-noise amplifier (LNA) with very low
power consumption. This LNA is a part of the bio-signal acquisition system described above.
When a sensor first detects a signal, the signal is extremely weak to perform subsequent process.
Thus, an LNA needs to amplify the signal without adding significant noise. Most importantly,
the LNA should consume as little power as possible to maximize lifespan of the medical device.
5
Our LNA is expected to consume less than 10µW and can minimize its power consumption
depending on how the data is going to compressed in the next module called compressed sensing
block.
Section 2.2 explains our business strategy focusing on industry analysis, market analysis,
and relevant trends that may affect the strategy to productize our product. The paper identifies
the industry and market to figure out the best way to present the product in the world. With
regards to commercialization of the capstone project, we take into account various trends to
sharpen our strategy.
2.2. Identification of Industry
We are mainly engaged in the activity of designing an LNA for ultra-low power and
long-durability applications. Our main target customers will be in the business of medical
implantable devices for real-time monitoring. We identify ourselves as the industry of
semiconductor & circuit manufacturing and design based on the IBIS World Industry Report
(Ulama). Since the industry has various facets, we need to identify ourselves in a more specific
area of the industry. We specify our area as designing integrated microcircuits.
In the integrated circuit industry, there are three main types of companies. The first type
is a company that both designs and manufactures microcircuits, such as Intel Corporation or
Advanced Micro Devices AMD. Also, there is a company that provides manufacturing services
to other companies, often referred to as a foundry, such as Global Foundries or Taiwan
Semiconductor Manufacturing Corporation TSMC. The third type is an entity that only designs
ICs, not actually manufacturing chips. The term, fabless, is used to describe a company that does
not own a foundry for the production of wafers (Ulama). The emergence of fabless is due to the
6
increasingly high capital barrier in the fabrication segment of the semiconductor industry (Chen).
The fabless business model requires much lower initial capital barrier and has higher return of
invested capital.
The main activity of the fabless is using a design platform provided by Electronic Design
Automation EDA companies. Cadence Design System is the most popular platform for designing
IC. We design a circuit schematic and simulate for its functionality and performance. The circuit
schematic is only an abstract level of circuit representation. It is not the actual blueprint of device
manufacturing. The next procedure is to design the layout of the device. The layout includes
more detailed information of how the microcircuit device is manufactured. The layout file is
what needs to be transferred securely to the foundries so the deisgn can be fabricated and sent
back to the designers for post-silicon testing. The procedure is often referred to as tape-out.
Since we mainly target our product application at the medical implantable devices, the
performance of the medical devices industry also greatly affects the profitability of our product.
Our potential customers will come from this industry. This industry includes manufacturers of
electro-medical and electrotherapeutic apparatuses, such as magnetic resonance imaging
equipment, medical ultrasound equipment, pacemakers, hearing aids, electrocardiographs and
electro-medical endoscopic equipment (Porter). The revenue of this industry has seen steady
growth of around 5 percent per year over the past 4 years. This makes our industry more
lucrative.
2.3. Potential Impact to the Industry
Our main goal is to provide a reliable solution for remote health monitoring. Patients with
chronic illnesses have to visit the hospital on regular basis; also this inconvenient practice may
7
overlook short-term variability of the patients’ health status. Currently the main problem with
remote real-time monitoring devices is power dissipation. In order to make these devices last
longer, conventional circuit design techniques are impractical. For these applications, special
circuit design methods need to be employed to suit the purpose of ultra-low power dissipation.
In electronic sensing devices, analog components consume much more energy than
digital components. In our case, the digital module on which we will deploy our design is able to
operate at as low as 22nW per channel (Allstot). However, the LNA, which is the interface
between the sensor and this digital module, can only operate at as low as microwatts of power
dissipation. Therefore if we can bring to the table a design that reaches the sub-microwatt region,
implantable devices for remote health monitoring are much more practical.
2.4. Threat of Alternative Technology
External health monitoring is the most threatening alternative technology that we are
facing. With the advent of more compact and powerful mobile platforms, wearable devices are
booming in performance as well as popularity. It is also a much safer application compared to
surgical operations that is a required step for implantable healthcare devices.
However the advantage of implantable devices should not be neglected. Implantable
devices guarantee more accurate results compared to external sensing. External monitoring is
prone to environmental variability such as humidity and temperature. Implantable devices are
more reliable because human bodies tend to maintain temperature and humidity. Implantable
monitoring devices can also interact with other implantable medical devices, such as
pacemakers.
8
2.5. Competition in the Industry
The wireless medical device market is at the verge of experiencing rapid growth in the
near future. In addition, we are not in the position where the dependence on the suppliers has
much force on the profitability of the industry. Our suppliers are EDA companies that provide IC
and layout design platforms. That is, in terms of a fabless, we do not need to worry much about
the manufacturing service, or the “foundry”, which is far more capital intensive and supplier
dependent.
The driving force here is rivalry. Many large companies are beginning to pay more
attention to the wireless medical device market, such as Analog Devices ADI, Qualcomm, and
Philips. The core of competition is expected to be the R&D of the technology and innovative
approaches to resolve the issue of long-lasting miniature device suitable for implantation.
Because the industry is demanding and tends to favor more advanced technology, the
profitability of the industry is undermined.
The main strategy that we adopt to counteract the threat from rivals is to focus on novel
design techniques. For instance, Texas Instruments is taking effort to improve the power
consumption of its DSP products to suit the purpose of ultra-low power applications (Ulama).
However, in a severely energy constrained sensor, Compressed Sensing eliminates the need for
digital signal processing (Allstot). New design methodologies tend to undercut the advantage of
established rivals such as TI and ADI, who take advantage from their well-established analog
design solutions.
In conclusion, the main tactic to tackle rivalry — the main threat in our go-to-market
strategy — is to aim at new and unconventional design techniques for ultra-low power
applications.
9
2.6. Potential Buyers, Suppliers, and End Users
As we have mentioned briefly in the section 2.1, our buyers are wearable or implantable
medical device manufacturing companies. We are targeting these companies because their
products could benefit from ultra-low power LNAs to maximize their lifespan. It will be a
laborious task for a user to charge or replace battery everyday. If a device is implanted in a
person’s body, changing a battery frequently is not a feasible option. Consequently, the
companies value electronic components with low power consumption. Our product will be one
of their interests. We target the potential buyers of our LNA at the medical devices design and
manufacturers, and they can integrate the chips with other parts of medical devices.
Aside from our potential buyers, we also need to identify who our end users are. Our end
users are people who are willing to utilize medical devices employing our LNA design. The
primary target-users would be patients suffering from chronic diseases. In their cases, real-time
monitoring is a great benefit in order to prevent life-threatening emergencies. We predict that
health-conscious people are also our potential end users because there has been a prevailing trend
of attention to healthy diet and lifestyle. Some wearable medical devices are able to record
heartbeat and blood pressure. People who want to keep track of their health status could benefit
from those devices.
There are two main suppliers for our product: Computer-Aid Design software company
and semiconductor foundry. Specifically, we need a subset of CAD software called Electronic
Design Automation software to design LNA circuits. As we have mentioned in section 2.2, we
will adopt fabless manufacturing model to produce LNA chips. In other words, semiconductor
foundry is required to provide service of fabricating our chip design.
10
2.7. Power of Buyers and Suppliers
According to Michael E. Porter’s The Five Competitive Forces That Shape Strategy, it is
essential to analyze power of buyers and suppliers and reshape them in our favor to be
competitive in the market (Porter). Fastidious customers can influence on the product’s price and
induce hyper-competition against rivals. On the other hand, powerful supplier can limit our
profitability by charging higher price.
In our case, power of buyers is moderate. The United States medical device market is
currently the largest in the world with the size of $110 billion (PRNewswire). It is expected to
reach $133 billion by 2016 (PRNewswire). In addition, there are almost 6,500 medical device
companies in the United States. The reason why the market has been growing rapidly is that
more people are eager to buy medical devices to treat chronic diseases. For instance, Emily Krol,
health and wellness analyst, says that there is a large market of consumers for products and
services specifically aimed to treat diabetes (PRNewswire). Although there are a plenty of buyers
in the market, the other factor makes the power of buyers strong. An LNA chip, which is a part
of semiconductor chip, is standardized and there are many competitors in the world. Thus,
buyers have few switching costs.
In contrast, supplier power is almost negligible and cannot affect our marketing strategy.
Buying CAD software from the company is merely one-time expense. After the purchase, we can
keep using the software to design circuits. The other supplier, semiconductor foundry,
experiences high competition in its own field, and the process of IC manufacturing and
fabrication is a standardized process (Ulama). As a result, we can switch easily among vendors,
weakening the power of suppliers.
11
2.8. Basic Marketing Strategy
Marketing mix is a fundamental business tool to sharpen marketing strategy. It is
associated with the four P’s: product, price, promotion, and place. Our product has to have the
advantages on power consumption and chip area over other competitors. Our main marketing
target would be medical device company that demands such specifications. Without
differentiating technology, it would be difficult to find buyers. In order to mitigate the power of
buyers, we need to make sure that our LNA chips could outcompete alternative products in the
market. In other words, if our chips can make a medical device that lasts twice longer than other
chips, our position in the market could be secured. We focus on wearable or implantable medical
device segment because this segment is sensitive to power consumption.
Price of our products should not be high unless there is a significant technological
advantage over alternative products. There is no guarantee for the affluence of our potential
buyers and end-users, thus setting a high profit margin may hamper the sale of our product.
Medical device companies will try to minimize the cost for producing their products so as to
improve market share of their products. In addition, the cost of the IC is subject to the effect of
scale, which means the cost for each product is reduced as the scale of production increases.
Therefore an effective approach to cut-down the cost of our product is to seek for larger market
share.
Since our main type of transaction is business-to-business B2B, there are not various
options for promoting the products. As a new entrant, our primary strategy is to create a solid
network with our partners in the business. To prove the product’s reliability and technological
advantages, we should hand out the samples of our LNA chips to the potential customers for free.
12
Although this strategy may weaken our financial status in a short term, it will help us secure the
customers in a long term.
Because we deal with B2B transactions, there is no need for physical stores for
distribution. The natural option is to conduct promotion campaigns directly targeted at
enterprises. This strategy implies that the promotion and quality of the products are very
important factors for our sales. As we expand the business, we may sell the products to different
markets other than medical device market to increase profitability.
2.9. Social Trends
As discussed in section 2.7, the U.S. society tends to pay more and more attention to
health care. And the Death: Final Data of 2013 reports the diseases in heart the dominant factor
of death among diseases (Center for Diseases Control and Prevention). This fact indicates the
mitigation of the buyers’ force, according to Porter’s five forces theory (Porter). Our direct
customers are wireless medical device companies, who will embed our design in their products.
We identify the end-users of our product as patients who are using these implantable devices. In
addition, people are becoming increasingly conscious of their well-being, which implies that
there is a great potential market for real-time health monitoring. In fact, CNBC is reporting a
rapid growth of wireless medical device in the United States.
2.10. Technological Trend
As mentioned in previous section, social trends are opening a potential market for realtime health monitoring market. In addition, the rapid development of Information technology IT,
such as the advances in video communication, wireless connectivity to the Internet, and the
13
increasingly popular web-assisted self-learning, is contributing greatly to the feasibility of
exploiting this market.
Meanwhile, what plays a more important role in our product is low power IC design and
manufacturing. Moore’s Law, which was proposed in 1965, predicted the roadmap of the
technological and economical advance of Integrated Circuit industry (Schaller). However, as the
semiconductor industry proceeds deeper into the submicron region, it is increasingly difficult for
the industry to advance the technology at the pace that the Moore’s law had successfully
forecasted for the past 40 years.
Since leakage power is becoming the dominant aspect of power dissipation in IC
components, low power design technology is becoming the main concern during contemporary
IC design. The influence of this industrial trend is both challenging us and exposing us to
opportunities. More competitions might be generated from academia due to research but we can
also make use of lately published papers to help with our low power design. In other words, the
field is moving forward and refreshing itself quickly and we are traveling in a technolgy
highway, exposed to both risks and opportunities.
2.11. Economic Trends
As we have discussed in the industry section, Integrated Circuit industry has high capital
barriers for the fabrication segment, though fabless companies experience relatively lower
barriers. Economic aspects play important roles in the industry. The Gross Domestic Product,
GDP, growth rate in US has also increased in the past years (The World Bank). And the
termination of Quantitative Easing by the US government is attracting a large amount of hot
money or refugee capital from the rest of the world to the US, which is good news for new
14
entrants in this industry. Combined with the fact that the semiconductor industry is reaching the
physical limitation of how small a device can be made, R&D investment in unconventional IC
design techniques are attractive to investors targeting at new applications such as the wireless
medical devices market.
2.12. Regulatory Trends
The U.S. government has well developed regulations in health care field. In the U.S.
regulations on electronics targeted at the medical industry and automotive industry has been
strict, thus we should be extremely cautious with our eventual implementation such that the
performance of our prototype tally with the regulatory specifications. Also, since we are design a
product that would be implanted into an individual’s body and collect data for a long time, new
regulations about this might appear in the future, which has not been seen. Regardless of this
factor of potential change, regulatory factors are negligible in our business.
2.13. Business Strategy to Cope with Contemporary Trends
As mentioned in the previous section, social and economic trends are favoring our
business, regulatory factors do not have strong influences on us, and the rapid development of
the core technologies relevant to the industry is putting us into both risks and opportunities.
Our strategy on these trends can be concluded as follow. We need to catch the opportunity
provided by the rapid growth of the wireless medical device industry and the recent uprising
trends of US economy, which means we need to act quickly. More importantly, we have to do
well in the technology part to avoid being outcompeted by rivalries in the industry. On the
technological aspect of the strategy, we should explore on novel design methodologies to gain
15
technological edges over competitors. Thus, keeping firm connection with academia is also a
crucial concern in our strategy.
16
3. Intellectual Property Strategy for Ultra-low Power LNA
3.1. Introduction
In July 1959, Robert Noyce was granted the first integrated circuit patent for his
innovative idea, and the actual functional prototype was fabricated in May 1960 (Computer
History Museum). After his patent, the number of patents related to IC research and design has
grown exponentially over the past half century. Protection of intellectual property is a sensitive
issue in the realm of IC research and design because excessive time and resources are required to
make a significant breakthrough. In reality, the situation of IP protection in the industry is
unsatisfactory. SEMI has reported that over 90% of the semiconductor R&D, design and
manufacturing entities have experienced certain extent of IP violations (SEMI). To avoid such
situation later on, we will discuss the patentability of our design and scrutinize other alternatives.
3.2. Patentability of Our Design
An Integrated Circuit is a product that includes at least one active element, in which
majority of the components and interconnects are formed on a single piece of material to perform
a desired function (World Intellectual Property Organization). Obviously the property of our
product matches with this definition, thus we should attempt to patent our design as an IC
product.
First, it is beneficial to clarify key technologies of interest in our project before
discussing the patentability of our design. The key features of our design include two aspects:
ultra-low power consumption and high signal-to-noise ratio. To achieve ultra low power design,
we will employ techniques to reduce the DC bias current of the circuit, such as sub-threshold
17
design. To achieve the goal of high signal-to-noise ratio, we will use techniques such as chopper
stabilization method and low-pass filter.
Here, we discuss some first order criteria for determining patentable ideas related to IC
design. These criteria include usefulness of the design, novelty of the concept, the extent of
innovation or differentiation compared to other designs aiming at the same functionality, the
extent of clarity of the design description, and the boundary of the scope of the definition
(Bellis). There are plenty of applications that we can employ our low-noise amplifier (LNA),
such as digital interface to the external off-chip components (Enz 335). Thus if that we can meet
all the required specifications, the usefulness of the design is unlikely to be contested. However,
we are utilizing several existing LNA design techniques to design our prototype, so the extent of
differentiation among counterparts is unlikely to be high. Therefore, we can conclude that it is
unlikely for us to patent our design.
3.3. Existing Similar Patent
It was difficult to find a patent that is closely related to our design. The main reason was
that we combined existing technologies to build up our LNA. After searching rigorously, we
found a patent named Mobile Wireless Communications Device Including A Differential Output
LNA Connected to Multiple Receive Signal Chains (U.S. Patent). It was published in 2010. In
the semiconductor industry, it is better to pick up a patent published recently to compare because
the technology life cycle tends to be short (Rethinking Patent Cultures).
18
3.4. Possible Similarity and Distinction
The patent is about a whole communication system, including receiver, transmitter,
antenna and other electrical components. In contrast, our LNA is merely a part of a bio-signal
acquisition system. The reason why this patent may be related to our product is that the system
contains an LNA as one of its components. As shown in the figure 1, the patent has an LNA that
takes the output of a receiver as input and delivers the output signal to multiple signal chains
through a power divider. However, our LNA is designed to take the output of an ECG sensor as
the input and deliver the signal to the compressed sensing module.
Figure 1 Schematic of the Depicted Patent
Although there is a certain extent of similarity with our system, we do not think that there
is a major overlap; there is almost no risk of developing our product without purchasing a
license. The patent is about a whole system, and it contains a component whose functionality is
similar to our product. The idea that is patented is focused on the whole functionality of system,
while we are more concentrated on designing the specific LNA component as part of the system.
Even the system’s LNA specification is different from our product. The specification of
our product requires only a narrow bandwidth, low carrier frequency and ultra low power
consumption. On the other hand, a mobile communication system operates at radio frequency
19
with a relatively wide bandwidth, and the power consumption is not necessary low, as is desired
in our application. When we design circuits under different specifications, the design approaches
and the outcome can be drastically different. Therefore a patent is conceived unnecessary for our
purpose.
3.5. Alternative Options and Potential Risks
Since applying for a patent is not feasible in our current situation, other alternatives
should be scrutinized carefully. There are four other ways to protect intellectual property from
external usage: a trademark, copyright, trade secret, and open source (Intellectual Property Crash
Course). All of them have different characteristics from a patent, and each one has its own
strength and weakness.
The first option is applying for a trademark. A trademark is simply a word or symbol that
embodies the source or sponsorship of a particular product or service (Intellectual Property Crash
Course). Our final product will be an LNA, and it does not have any particular name or symbol
to strengthen its marketability. Also, a trademark does not protect the essence of our project,
which is the approach to tackle the challenges in low-power LNA design.
The Second option is copyright. Unlike patent, copyright only protects the expression of
an idea, and the idea itself is not secured (Intellectual Property Crash Course). Most importantly,
in the United States, while IC layout design is commonly protected by patent, circuit schematics
are not commonly protected (U.S. Copyright Office). Since our main focus is not the layout
design, this option is not appropriate.
The third option is trade secret. The recipe for Coke is a classical example of the trade
secret approach. Unlike patent, a person can keep a trade secret for a long time. A person does
20
not need to worry about other people seeing how his product works. A trade secret is perfect if a
product is useful for several generations. However, in our case, the lifespan of the technology is
very short, which is a fundamental limiting factor of the effectiveness of this approach.
The last point is an open source. This option is meaningful if we deal with software. The
whole purpose of making a product open source is to improve the product with collaborative
effort. In our case, the project is a circuit design so that this option is also not applicable. While
other professionals in the field could have beneficial suggestions or criticizes a certain design, it
is extremely difficult to contribute to another person’s design by improving based on the design.
Finally, we should think of some potential risks of not protecting our ideas with legal
enforcement. We conceive that, at this point, there is no such urgent need. Usually it takes years
for a patent to be issued after the first application. It means that we need to devote extra time and
resources to such process. Knowing that technology life cycle is very short in our industry, it is
better for us to focus on developing more superior technology until it is more suitable for a
patent.
21
4. Technical Contributions
4.1. Overview
The overall objective of our Capstone Project is to design a Low Noise Amplifier (LNA)
for ultra-low power wireless transceivers in bio-signal acquisition systems. The LNA has mainly
two features: ultra-low power dissipation and high noise tolerance, while achieving suitable gain
for further processing.
The project consists of mainly three parts:
•
Designing and analysis of the chopper stabilization modulator and demodulator;
•
1/f flicker noise analysis; amplifying circuitry design;
•
Analysis of the non-linearity and distortion analysis of the amplifying circuit;
•
Thermal noise analysis and techniques to compensate for thermal noise in low power
LNA designs.
The task divisions are in bundles aiming at a specific section of the design problem. This
makes it easier for one person in the team to concentrate on the tackling a specific problem in the
design. When each individual task is completed, the design is merged to test for systematic
functionality and performance. This is the ideal case while in reality the segmentation of the
analog / RF circuit design is non-trivial since the various specifications may imply opposite
directions in the design. For instance, to eliminate the effect of the flicker noise, the chopping
frequency should be high, while for thermal noise the modulation frequency should be low. Thus
co-optimization is often required.
In the project I am mainly engaged in the design of the amplifying circuit and distortion
analysis. My component is the main amplifying component in the design. It is also crucial for the
22
amplifier to have reasonable gain at frequency range from mid-band to -3dB band of a few kHz.
It is also crucial to keep the non-linearity of the circuit low to alleviate the impact of distortion.
4.2. Literature Review
With the advent and rapid development of the solid-state device and integrated circuit
fabrication technology, it is possible to achieve very high voltage gain with minimal static
power. But the ultra-low power design has a fundamental limiting factor, which is the electronic
noise inherent to the active devices as well as the passive resistors in the system (Leacher 1515).
The existence of electronic noise and the non-linear response of solid-state devices are the main
sources of distortion in the LNA design. Thus reducing the gain in the power of the inputreferred noise is beneficial for reducing the distortion of the amplifier.
Our LNA design is part of the design project for an ultra-low power wireless medical
implantable transceiver for bio-signal acquisition system, based on compressed sensing
technique. A structural diagram of the system is shown below, where ECG is the ECG sensor
module that senses the electro-cardio signal of the person implanted with this device. The LNA
is the focus of our Capstone Project. CS is the compressed sensing module, which further
processes the output signal of our LNA design. ADC is the analog-to-digital converter that
converts the compressed signal into digital signal. The power amplifier and the antenna then
transmit the modulated signal to a portable mobile base station (Dixon 156).
Figure 2. Structural Diagram of Bio-signal Acquisition System
23
We based our design on the SAED 30nm technology provided by Synopsys (32/28nm
Interoperable PDK). This differs from the preceding study in that the technology we used has a
supply voltage of 1.05V, which is lower than the referenced study of 1.3V (Lim 203). Although a
reduction of supply voltage helps alleviate power concerns, the devices from this technology has
serious short-channel effects that hamper the gain as well as the frequency performance of
amplifier circuits comprised of these devices. Also since the operation regions of the transistors
are not well defined in the deep-submicron region, the approach for sizing the transistors also
differs from the long-channel devices.
The flicker noise of MOSFET devices is given by the equation below:
i12 f =
K f I D Δf
L2Cox f
where Kf is the process-dependent coefficient, ID is the DC bias current Cox is the gate
capacitance, and L is the length of the MOSFET. Since the frequency of the ECG signals, i.e. the
heart rate of a human is typically in the range of 0.1 to 10 Hz, the ECG signal is extremely prone
to contamination of flicker noise (Lin 680).
4.2.1. Common-Source Fully-Differential Amplifier
A design approach to incorporate chopper stabilization technique with a fully-differential
amplifier. The chopper stabilization technique is applied both at the input and the output of the
amplifier to reduce the amount of contamination to low frequency signals by flicker noise. The
schematic in figure 3 presents the depicted topology. The mid-band gain of the amplifier is
Av =
24
gm1
gm 4
where gm1 is the transconductance of the input differential pair and 1/gm4 is the approximation of
the output impedance due to the diode connected PMOS load (Enz 338).
Figure 3. Fully Differential Amplifier (Enz 338)
4.2.2. Ring-Amplifier
Conventional analog circuit design approaches is becoming less effective since the
scaling of technology strongly favors the time domain of high-speed digital circuit, thus a highly
scalable approach for deep sub-micron analog design should be found in the time domain
(Hershberg 2928).
A novel approach to design the amplifiers in the deep sub-micron technology nodes is the
ring-amplifier topology. Ring-amplifier is an unconventional way to achieve high gain in deep
sub-micron amplifier design, since it utilizes the large signal slew-based operation of three-stage
ring-oscillators (Hershberg 2929). The structure is particularly attractive in contemporary deep
sub-micron technology because of its relatively simple structure and its capability to suit
technology scaling. However, ring amplifier design is a relatively new topic and there has been
25
no analytical approach given by researchers yet, the design of ring-amplifier requires a certain
extent of intuition and fine-tuning.
The topology of the ring-amplifier is presented in figure 4. The fundamental difference
between the ring-amplifier and its origin, the ring-oscillator, is that the signal path for the third
stage inverter is split into two different paths for the PMOS and NMOS transistors respectively.
This enables the input of the pull-up transistor and the pull-down transistor to be embedded with
separate DC voltages, thus enabling both transistors of the output stage to operate at the subthreshold region. Biasing the output transistors at sub-threshold regions is not only beneficial for
power consumption consideration, but also crucial for the biasing the amplifier at appropriate
operation regions, particularly stability of the closed-loop configured ring-amplifier. The
difference between the embedded DC bias voltages VRN and VRP is defined as the dead-zone
voltage VDZ. In order to stabilize the amplifier, the dead-zone voltage should satisfy the
following relationship:
VDZ = VRN − VRP ≥
VDD − VSS − 2VT
A2
where A2 is the voltage gain of the second inverter stage. Aside from the DC bias voltages, the
input range is also a crucial concern for the stable operation of the amplifier. When applying
MDAC feedback network structure, the input DC level is not a key concern since the sampling
capacitors between the input terminal and internal node can hold the DC voltage difference; in
order to stabilize the amplifier, the AC input magnitude should satisfy the following relationship:
VIN ≤
⎞
1 ⎛ VDD − VSS
− 2VOS ⎟
⎜
2A1 ⎝
A2
⎠
26
where A1 and A2 are the voltage gains of the first two stages respectively, and VOS is the DC
offset voltage for the input of the second stage. In bio-signal applications the AC input range is
not a serious threat to the amplifier operation because of the small signal amplitude, but DC
input voltage may vary significantly, which is compensated by the switched capacitor nature of
the pseudo-MDAC structure that periodically corrects the DC bias points (Carnes 5251).
Figure 4. Schematic of the basic ring-amplifier structure
The time-domain response of the ring amplifier, once the ring-amplifier satisfies the
stability criterion, is categorized into three periods: initial ramping, stabilization and steady
amplification (Hershberg 2930). When the amplifier transitions from the periodic reset mode into
the amplifying mode, the internal nodes in the amplifier will experience initial slewing due to
physical mismatches and external interferences. This initial voltage ramping will create an
overshoot voltage, which in the case of a ring-oscillator makes the successive slewing gain a
larger overshoot voltage and eventually achieves oscillation. The reason why the ring-amplifier
does not oscillate is that the overshoot voltage for the output transistors induced by the
preceeding opposite overshoot voltage is significantly smaller than the ring-oscillator case, since
27
the output transistors are biased at subthreshold region. Thus the output resistance is significantly
larger once the output enters the stable operating regions. This fundamental difference reduces
the overshoot voltage of each successive voltage ramping and by a number of repetition the
output voltage will stabilize and lock itself into the dead-zone region. The differences between
each oscillation is
ΔVovershoot =
t d ⋅ I slew
∑ Cout
Thus the expression for stability criterion is presented below.
⎞
t d ⋅ I slew
1 ⎛ VDD − VSS
≤
− 2 ⋅VDZ ⎟
⎜
∑ Cout 2 ⋅ A1 ⎝
A2
⎠
This equation gives a clear relationship between td, Islew and VDZ, which link to the speed,
output range and power dissipation of the amplifier. The effect of feedback network on stability
and gain will be discussed in detail as the design process in section 4.3.
Figure 5. Ring Amplifier Applied in MDAC Feedback Structure (Hershberg 2929)
28
4.3. Methods and Materials
4.3.1. Ring Amplifier Design Process
The motif of designing the LNA based on the ring-amplification structure is due to its
attractiveness for the purpose to achieve high gain and consuming low power at ease due to its
relatively simple structure, as long as the stability issue for the amplifier is appropriately
resolved. Unfortunately, the idea of using ring-oscillator structure to implement amplifier is a
relatively new idea, thus there has been no readily available analytical methodology to describe
the slew-based amplifying operation of the amplifier in the frequency domain. Thus, much of the
effort we have made during the design process of ring-amplifier is explorative, through which we
identify some of the important influencing factors for the ring-amplifier performance, as well as
some issues we encountered during the design process.
The main difference between using an inverter as an amplifying stage and a conventional
common source amplifier with active load device is that the signal path is no longer connected to
only one of the transistors in the stage, enabling the push-pull operation of the NMOS-PMOS
pair which is similar to the operation of static CMOS logic gates. In general, the voltage transfer
curve of the inverter transitions more rapidly as channel length of the transistor increases. Also,
larger W/L ratios for the transistors will increase the gm of the transistors, which increases the
gain of the inverter.
First, the allocation of gain for each stage is determined. Since the gain of each stage
affects the noise performance of the original ring-amplifier, i.e. before the addition of noisealleviation techniques such as the chopper technique, it is important to appropriately assign the
gain of each stage. The overall noise figure of the amplifier is defined as:
29
⎛
F − 1 F3 − 1 ⎞
NFRING−AMP = ⎜ F1 + 2
+
A1
A1 ⋅ A2 ⎟⎠ dB
⎝
where the subscript depicts the order of the inverter stage. Thus for a given gain requirement,
allocating the highest gain to the first stage will help significantly reduce the noise figure of the
amplifier. Thus the lengths of the transistors for the first stage should be set large to achieve high
gain. Also, since our scenario application is ECG sensors whose output voltage is typically less
than 1mV, if the target gain for the successive two stages is 46dB than the output swing of the
first stage need not be larger than a few millivolts to achieve an output swing of 300mV.
Therefore the static current of the first stage could be reduced to conserve power. For the second
stage, the DC bias voltages of the second stage greatly determines the DC voltage of the third
stage, which is crucial for the stability of the inherently closed-loop configured ring-amplifier.
Since the DC bias voltages VRN and VRP generate the DC voltages VBP and VBN through 2nd stage
inverters, VBN and VBP should be stable against variations in VRN and VRP. One method is to size
the transistors to have maximal threshold voltage. However, this is difficult to practice since
once the ring-amplifier settles to the operating point, the second stage inverters are prone to
entering triode region, resulting in low overall voltage gain (Lim 202). A more practical way to
stabilize VBP and VBN is to reduce the large signal gain of the 2nd stage, thus the lengths for the
transistors of the 2nd stage should be minimal.
The sizing of the 3rd stage inverter is the trickiest part of the design process, not only
because it affects the overall gain of the amplifier, but also the stability of the amplifier. The
transistors in the 3rd stage should be biased in the sub-threshold region by VBP and VBN, in order
to trigger the diminishing overshoot transition, as discussed in section 4.2.2. Therefore it is
appropriate to size the transistors such that they possess the largest Vt values. This will result in a
30
large dead zone region, which alleviates the dependency of stability on VBP and VBN variation.
Once the threshold voltages of the amplifier is determined, the values for VBP and VBN can be
determined, as well as the values for VRP and VRN. While the determination of VRP and VRN can
be easily done by simulating the VTC of the 2nd stage inverter, the determination of VBP and VBN
is not supported by a convenient model yet. The sub-threshold conductance due to the underlying
parasitic bipolar transistor is
I ds = 100 ⋅
W − q(Vgs −Vt )/ηkT
⋅e
(nA)
L
which clearly indicate an exponential increase with increasing Vgs (Hu 283). In a normal
common source amplifier this equation is sufficient to derive the gm and ro of the amplifying
transistor, but since in an inverter there is no load device and both transistors collaborate to
amplify the signal and ro is constantly varying, this approach is inaccurate due to the push-pull
operation of the two complimentary devices. The third stage is also crucial for the stability of the
amplifier in closed-loop configuration, so the 3rd stage also needs to be adjusted to compensate
for charge sharing due to Cf. The schematic of the closed-loop ring amplifier is shown in figure 6.
Figure 6. Ring-Amplifier Schematic
31
The next major concern in the ring amplifier design is stability. Because of the inherent
closed-loop configuration, the stability of the amplifier not only depends on the three stages, but
also depends significantly on the feedback factor, i.e. mainly the feedback capacitor Cf. For biosignal applications the bandwidth requirement for the amplifier is relaxed, therefore in our design
we sacrificed bandwidth for stability of the amplifier by introducing large capacitors as the
charge holding capacitors in the input to create a dominant pole at the input. Reducing bandwidth
also helps us to meet with the stringent power requirement imposed by implantable applications.
The feedback factor is
f=
Cf
C f + Cin
Since the amplifier gain is inversely related to the feedback factor, Cf should be as small as
possible. However the pole at the output node also depends on Cf, thus its value should also be
large enough to stabilize the amplifier.
In the process of the ring-amplifier design, we discovered that the closed-loop gain and
stability are not significantly affected by the inverter stages, but significantly affected by the
feedback loop. This is mainly because the three slew-based inverter stages render the amplifier
high open-loop gain; plus at low frequency range, capacitors exhibit high impedances and
dominate the feedback factor. In our design process, the feedback capacitor has a value of Cf =
10fF. A capacitor with such a small value is not feasible in older technologies due to fabrication
mismatches, however in modern deep-submicron technologies, with careful layout matching
techniques small capacitor values can be well controlled. E.g. in a 32nm SOI technology, this
capacitor value can be implemented with multi-finger fringe capacitor of 0.45fF / finger
32
(Tripathi). The input capacitances are set to 10pF in order to have small feedback factor and
hence enlarge the closed-loop gain.
The switches in the circuit are implemented with simple NMOS transistors, since except
for the switch that is connected to the input terminal, none of the other switches are in the signal
path. The switches mainly influence the noise of the amplifier and spikes during the periodic
reset transitions. Shot noise and flicker noise are directly related to the DC current flowing
through the junctions in the transistors, thus the switches contributes little noise since little DC
current flow through them. Thermal noise is related to the on-state resistances of the switches,
therefore increasing the W/L ratios of the switches will reduce the thermal noise introduced by
the switches. In practice, however, we discovered that if the size of the switches is large, the
voltage spikes are large, due to charge injection effects induced by the parasitic Cgs and Cgd.
Combined with the fact that the noise contribution of the switches is relatively minor, the
switches are minimum sized (Gray 736).
4.3.2. Adjustments and Improvement
There are several problems with the original circuit topology presented in figure 6. First,
the stability of the amplifier is extremely sensitive to variations of the DC voltages VRN and VRP,
Cf and DC input voltage. In extreme cases, 1mV variation of VBP and VBN will make the ringamplifier oscillate. Second, there is significant DC output voltage shift when in the amplifying
mode. This DC shift is mainly due to the DC shift of the first stage output and therefore inert to
adjustments in transistor sizing and DC bias of the third stage. Increasing the value of the input
capacitors of the second stage CDZ will alleviate the output DC shift, but only by as much as 10%.
33
Another problem is the transistor sizing of the first stage. As discussed in section 4.3.1, the gain
of the first stage should be high. However the gain of the first stage affects almost all the aspects
of the ring-amplifier performance significantly, including stability, gain, and output DC level,
thus modifying the gain of the first stage will very likely make the ring-amplifier inoperable.
Another factor that makes the previous topology inappropriate for our scenario
application is its single-ended input structure. The DC voltage level from the ECG level can vary
by as much as 100µV, thus many ECG sensors provide differential output to enable amplifying
schemes that are inert to input DC voltage shifts (Texas Instruments 3). Furthermore, since the
output signal range is at low frequencies, the primary type of noise that contaminates the desired
signal is the flicker noise. The flicker noise cancellation scheme we apply to the amplifier is the
chopper stabilization technique, which also requires differential signal paths for optimal noise
cancellation performance.
To enable the incorporation of chopper stabilization technique with the amplifier, it is
required to adjust the input structure to differential input structure. A structural diagram of the
modification is shown in figure 7. Because approximately 49% of the total amplifier noise is
introduced by the 1st stage, applying chopper module to the 1st stage is sufficient for the purpose
of noise reduction. Since the 1st stage is now differential output, the corresponding input
structure of the 2nd stage is also changed into differential input. Another modification is to
directly apply the embedded bias voltages VRN and VRP to the input of the last stage, instead of
the input of the second stage. The rationale behind this modification is that the variation of VRN
and VRP by 1mV could result in as much as 60mV deviation for VBN and VNP. Thus the amplifier
is more stable if VRN and VRP are directly embedded to VBN and VBP, respectively.
34
Figure 7. Fully-Differential Structure of the Modified Amplifier
High 1st stage gain is desired for optimal noise performance. One approach is to use
folded-cascode structure. However folded-cascode structure is prone to performance
deterioration if biased with low DC current, not to mention it contributes much more noise due to
resistive degeneration. Thus a fully differential common-source amplifier is first adopted. While
having better noise performance than cascode amplifier, the gain of such structures in 30nm
technology could barely reach 26-28dB.
When we refer back to the high gain property of the inverter in deep-submicron
technology, we noticed that the input signal flows to both the NMOS and PMOS transistors,
resulting almost twice as much transconductance as the simple common-source amplifier case.
We also noticed that the amplifier ring-amplifier has been established with reset-amplify mode
of operation. These allow us to introduce a dedicated amplifying structure, which we call fully
differential inverter. The symbolic representation and the transistor-level schematic are presented
in figure 8. This structure improves the gain of the 1st stage to 32dB. Another advantage of this
structure is the utilization of a current source to determine the bias current of the differential
inverter. This effectively reduces the extent of output DC shift.
35
Figure 8. Dedicated Fully Differential Inverter
Figure 9. Schematic of the Modified Ring-Amplifier
There are prerequisites for implementing this structure. First, the input AC signal range
should be small enough to ensure both transistors are biased in saturation region. Second, there
should be a frequent reset action to ensure that the DC bias for the fully differential inverter is
corrected periodically. The environment of the 1st stage inverter satisfies both conditions. A
drawback of this structure is the increased noise level due to utilizing the transconductance of the
NMOS transistor, but since we would applying the chopper technique to the 1st stage, the noise
introduced by the 1st stage will be modulated by the chopping frequency, thus it is less likely to
36
contaminate the desired signal which is modulated back to low frequency ranges by the 2nd
chopper. The schematic of the modified amplifier is presented in figure 9. Also note that the
dominant pole is at the output node of the second stage. This is obvious in that the dead-zone
capacitors should be kept large in order to lower the frequency of the low-frequency zero, hence
enlarging the low-frequency voltage gain. Also the second stage has moderate gain, thus the
output resistance of this node is high.
Another point of notice is that since the LNA is to amplify very-low frequency signals,
the zeros in the circuit should locate at extremely low frequencies. The AC feed-through
capacitors CDZ between the second stage and the third stage should be enlarged, in order to move
the RHP zero contributed by CDZ to be located at extremely low frequency. To add stability to
the amplifier, a differential-mode configured frequency compensation capacitor is added to the
output node of the second stage. It further reduces the bandwidth of the amplifier and the phase
margin is increased to approximately 30° when the amplifier is open-loop configured.
4.4. Results and Discussion
Figure 10. Transient Waveform of the Ring-Amplifier
37
The transient simulation of the ring-amplifier is shown in figure 10. The input stimulus
has Vpp = 1mV and fin = 10Hz. The basic ring-amplifier structure also suffers from output DC
shift when the amplifier enters amplifying mode.
The small signal PAC simulation result is shown in figure 11. The low frequency voltage
gain is Av = 51.73dB. Due to convergence problems encountered in frequencies higher than
20MHz in the PAC simulation, the plot terminates at 20MHz; however the phase and voltage
gain magnitude given in the plot below indicates a positive phase margin.
Figure 11. PAC Frequency and Phase Response of the Single-Ended Ring-Amplifier
The output noise spectrum is shown in figure 12. At 10Hz the output noise voltage is
871uV/(Hz)1/2.
38
Figure 12. Output Noise Voltage Spectrum of the Single-Ended Ring-Amplifier
The output total harmonic distortion is shown in figure 12. At 1mV input voltage
magnitude the output THD is 14.3%.
Figure 12. Output Total Harmonic Distortion of the Single-Ended Ring-Amplifier
39
The transient simulation waveform of the modified ring-amplifier, i.e. the amplifier
without the choppers, is shown in figure 13. As can be seen from the waveform, the output DC
shift is significantly alleviated.
Figure 13. Transient Waveform of the Differential Input Ring-Amplifier
Figure 14. PAC Frequency and Phase Response of the Modified Ring-Amplifier
40
The small signal PAC simulation result is shown in figure 14. The low frequency voltage
gain is Av = 63.93dB. Similar to the single-ended case, a positive phase margin is expected.
The output noise spectrum is shown in figure 15. At 10Hz the output noise voltage is
1300uV/(Hz)1/2.
Figure 15. Output Noise Voltage Spectrum of the Modified Ring-Amplifier
The output total harmonic distortion is shown in figure 16. At 1mV input voltage
magnitude the output THD is 27.8%.
41
Figure 16. Output Total Harmonic Distortion of the Modified Ring-Amplifier
The transient waveform for the modified amplifier with chopper stabilization is shown in
figure 17.
Figure 17. Transient Waveform of the Differential Input Ring-Amplifier with Chopper Stabilization
42
The small signal PAC simulation result is shown in figure 18. The low frequency voltage
gain is Av = 57.77dB. By applying differential compensation capacitor at the output node of the
second stage amplifier, the UGB is brought down to below 20MHz where the PAC result would
converge. The PAC analysis gives UGB = 12MHz, and phase margin of approximately 30°.
Figure 18. PAC Frequency and Phase Response of the Modified Ring-Amplifier
Figure 19. Output Noise Voltage Spectrum of the Modified Ring-Amplifier
43
The output noise spectrum is shown in figure 19. At 10Hz the output noise voltage is
120uV/(Hz)1/2. Compared to the unchopped amplifier case, the total output noise voltage is
reduced by approximately an order of magnitude.
The output total harmonic distortion is shown below. At 1mV input voltage magnitude
the output THD is 28%.
Figure 20. Output Total Harmonic Distortion of the Modified Ring-Amplifier
The FFT analysis results of the original ring-amplifier, the unchopped and the chopped
differential input amplifier is shown in Figure 21. While having good noise performance, the
amplifier suffers from serious distortion. The distortion of the unchopped amplifier is well
controlled, but noise performance is inferior, due to the noise contributions of the NMOS
amplifying transistors of the 1st stage. The chopped amplifier has better noise performance
compared with the unchopped case, while compromising some extent of distortion.
44
Figure 21. FFT Analysis Results for the Three Types of Implementations
A comparison of the performance of the three implementations is given in Table 1. It can
be concluded that the fully differential ring-amplifier with chopper stabilization has the most
well balanced performance in terms of distortion and noise performance.
Table 1 Performance Comparison of Three Different Topologies
Topology
Original
Static Power
(µW)
2.75
ACL
(dB)
51.73
THD
(Vin=1mV) (%)
14.3
Noise Von (fin=10Hz) (µV/Hz1/2)
871
SNR
(Vin=1mV) (dB)
76.76
45
Unchopped
5.51
63.93
27.8
1300
60.33
Chopped
5.37
57.77
28
120
78.75
The design process of our Capstone Project has rendered us some insights in ultra-low
power LNA design as well as some useful experiences for designing ring-amplifier, which is a
relatively new idea. Here we conclude some of the important points of notice for ring-amplifier
design. For low noise purposes, the 1st stage should target at high gain. The second stage should
have gradual VTC curve to stabilize the DC bias voltage of the 3rd stage. The DC bias voltages
for the 3rd stage VBP and VBN should be co-optimized with the sizing of the transistors of the 3rd
stage, to stabilize the amplifier in closed-loop configuration. The closed-loop gain is greatly
influenced by the feedback factor, due to the large open-loop gain of the three-stage inverters.
46
5. Concluding Reflections
Through the Capstone Project, not only have I honed my knowledge and skills in low
power and low noise amplifier design at low frequency ranges, but also gathered a sense of how
to identify, address and resolve issues encountered when conducting a project. Initially, the main
problem we encountered was achieving high gain when designing with 30nm technology. We
have tried various conventional amplifier topologies but all failed to achieve the desired
performance. During the design process we discovered a relatively new topology called the ringamplifier, which is a competent candidate for achieving high gain while consuming ultra-low
power dissipation. Albeit its potential, the ring-amplifier is currently not thoroughly studied, thus
there is no convenient model to pilot the design process. Some key insights for the design of
ring-amplifier resulted from the exploration process, and are discussed in detail in the technical
contribution section.
Much of the plight we encountered is due to lack of analytical model to pilot the
procedure of ring-amplifier design. Thus it will be extremely beneficial not only to the ringamplifier design itself, but also to the realm of analog design in the deep-submicron technology
nodes, if a readily available analytical model for ring-amplifier is proposed.
47
Work Cited
"32/28nm Interoperable PDK." Synopsys. Synopsys Armenia Educational Department, 2011.
Web. 10 Mar. 2015.
Allstot, David J. et al, “Compressed Sensing Analog Front-End for Bio-Sensor Applications”,
IEEE Xplore Database, Web. 15 Feb. 2015
Bellis, Mary. “How to Qualify for a Patent.” About.com Inventors. About.com, n.d. Web. 24 Feb.
2015.
Carnes, Josh. and Un-ku Moon. “The Effect Of Switch Resistance On Pipelined ADC MDAC
Settling Time”. Circuits and Systems 2006 (2015): 5251. Web. 16 Mar. 2015.
Centers for Diseases Control and Prevention. “Deaths: Final Data for 2013”. Web. 15 Feb. 2015
Chan, Wen-Yen. “Mobile wireless communications device including a differential output LNA
connected to multiple receive signal chains.” U.S. Patent No. 8,125,933. 28 Feb. 2012.
Chen, Roawen. “The Future of Semiconductor Industry from ‘Fabless’ Perspective”, Web. 15
Feb. 2015
Dixon, Anna M., Emily G. Allstot, Daibashish Gangopadhyay, and David J. Allstot.
"Compressed Sensing System Considerations for ECG and EMG Wireless Biosensors." IEEE
Xplore. 156 - 166, Apr. 2012. Web. 9 Mar. 2015.
Enz, Chiristian C., Eric A. Vittoz, and F. Krummenacher. "A CMOS Chopper Amplifier."
Journal of Solid-State Circuits. IEEE Xplore, June 1987. Web. 9 Mar. 2015.
Gray, Paul R., Paul J. Hurst, Stephen H. Lewis, and Robert G. Meyer. 4th ed. New York: John
Wiley and Sons, 2000. Print.
Hershberg, Benjamin et al. “Ring Amplifiers For Switched-Capacitor Circuits”. 2012 IEEE
International Solid-State Circuits Conference (2012): 2928-2935 Web. 15 Mar. 2015.
48
Hoopes, Stephen. “IBISWorld Industry Report 52411b Health & Medical Insurance in the US”.
Web. 15 Feb, 2015
Hu, Chenming. “Modern Semiconductor Devices For Integrated Circuits”. Upper Saddle River,
N.J.: Prentice Hall, 2010. Print.
“Intellectual Property Crash Course.” Patents, Copyrights, Trade Secrets, & Trademarks. N.p.,
n.d. Web. 25 Feb. 2015.
Leach, Marshall W., JR. "Fundamentals of Low-Noise Analog Circuit Design." Proceedings of
the IEEE 82.10 (1994): 1515. Web. 9 Mar. 2015.
Lim, Yong, and Michael P. Flynn. “11.5 A 100MS/S 10.5B 2.46Mw Comparator-Less Pipeline
ADC Using Self-Biased Ring Amplifiers”. 2014 IEEE International Solid-State Circuits
Conference Digest of Technical Papers (ISSCC) (2014): 202-203 Web. 15 Mar. 2015.
Lin, Chia-Hung. "Frequency-domain Features for ECG Beat Discrimination Using Grey
Relational Analysis-based Classifier." Computers & Mathematics with Applications 55.4 (2008):
680-90. Web. 11 Mar. 2015.
McKinsey & Company. “McKinsey on Semiconductors”. Web. 15 Feb. 2015
Porter, Michael. "The Five Competitive Forces That Shape Strategy." Print.
Porter, Michael. “What is Strategy?” Harvard Business Review, Print.
PRNewswire. “Global Wireless Portable Medical Device (Monitoring, Medical Therapeutics,
Diagnosis, Fitness & Wellness) Market - Forecast to 2020.” Web. 15 Feb. 2015
“Role of Intellectual Property in the Intellectual Property (IP) Challenges and Concerns of the
Semiconductor Equipment and Materials Industry.” Role of Intellectual Property in the
Intellectual Property (IP) Challenges and Concerns of the Semiconductor Equipment and
Materials Industry (2008): 1. SEMI. SEMI, Apr. 2008. Web. 24 Feb. 2015.
49
Schaller, Robert R. “Moore's law: past, present and future.” Spectrum, IEEE 34.6 (1997): 52-59.
Texas Instruments. “Low-Power, 2-Channel, 24-Bit Analog Front-End for Biopotential
Measurements (Rev. B)”. (2011): 3-4. Sept. 2012. Web. 15 Mar. 2015.
“The Role of Patents and the Wireless Medical Device Market.” Rethinking Patent Cultures.
N.p., 06 May 2014. Web. 26 Feb. 2015.
The World Bank “GDP growth (annual percentage)” Web. 15 Feb. 2015
Tripathi, Vaibhav, and Boris Murmann. “Mismatch Characterization Of Small Metal Fringe
Capacitors”. IEEE Trans. Circuits Syst. I 61.8 (2014): 2236-2242. Web. 15 Mar. 2015.
Ulama, Darryle. “2014
IBISWorld Industry Report 33441a: Semiconductor & Circuit
Manufacturing in the US”. Web. 15 Feb. 2015
United States. U.S. Copyright Office. “UNITED STATES CODE TITLE 17—COPYRIGHT”.
N.p.: n.p., n.d. WIPO. Web. 25 Feb. 2015.
“World Intellectual Property Organization.” Washington Treaty on Intellectual Property in
Respect of Integrated Circuits. World Intellectual Property Organization, n.d. Web. 24 Feb. 2015.
“1959 - Practical Monolithic Integrated Circuit Concept Patented” 1959 - Practical Monolithic
Integrated Circuit Concept Patented - The Silicon Engine | Computer History Museum. N.p., n.d.
Web. 24 Feb. 2015.
Yang, Xiao. et al, “Design of Low Power Low Noise Amplifier for Portable Electrocardiogram
Recording System Applications”, IEEE Xplore Database, Web. 15 Feb 2015.
50